Circulating cooling water system period operation optimization method based on maximum net income of power plant

Abstract

The invention provides a circulating cooling water system period operation optimization method based on maximum net income of a power plant, which comprises the following steps of: establishing a heat transfer model of each component of a circulating cooling water system of the power plant, and performing comprehensive heat transfer calculation of the system; establishing a system operation water temperature database, and establishing a system water temperature black box model by adopting a deep learning method; a condenser pipe net scaling rate model is established, and the operation water temperature calculated by the condenser is corrected; establishing a power plant operation period maximum net income double-layer optimization model; solving the model to obtain a power plant circulating cooling water system adjusting period optimal division scheme, a condenser optimal cleaning scheme and a pump unit optimal operation scheme in each adjusting period; and connecting all annual operation periods of the power plant in series, establishing an annual maximum net income model of the power plant, and calculating and evaluating the annual operation optimization potential of the circulating cooling water system of the power plant. Energy and water can be saved for the condensing power plant, benefits are improved, and the effect is remarkable.

Description

Technical Field

[0001] The invention relates to the field of industrial efficiency improvement, energy conservation and emission reduction, and in particular to a time period operation optimization method for a circulating cooling water system based on maximum net profit of a power plant. Background Art

[0002] As the proportion of renewable energy power generation increases, traditional condensing power plants have heavy peak-shaving tasks, and the power generation load and ambient temperature change frequently, which makes it a challenge to optimize the matching of the operating conditions of the condenser circulating cooling water system and the peak-shaving operation of the thermal system. This has not been solved in actual engineering and theoretical research. When the power generation load requirements of the power plant change, most power plants only adjust the steam load of the steam turbine unit, and do not easily adjust the operation plan of the cooling water system pump unit. They still adopt the high energy consumption and large cooling water flow operation mode before peak regulation, resulting in poor power generation performance of the steam turbine unit and serious energy waste of the cooling water system. The existing technology can maximize the instantaneous net power generation of the power plant by optimizing the operation plan of the circulating cooling water system, but due to changes in ambient temperature and required power generation load, this method does not consider the adjustment of the operation plan of the circulating cooling water system pump unit and cannot be implemented in engineering. Summary of the invention

[0003] Purpose of the invention: In order to overcome the shortcomings of the background technology, the present invention discloses a method for optimizing the period operation of a circulating cooling water system based on the maximum net profit of a power plant.

[0004] Technical solution: The method for optimizing the period operation of a circulating cooling water system based on the maximum net profit of a power plant disclosed in the present invention comprises the following steps:

[0005] S1. Establish heat transfer models for each component of the condenser circulating cooling water system of the power plant and perform comprehensive heat transfer calculations for the system;

[0006] According to the condenser heat transfer balance equation, the following equations are listed to solve the condenser heat transfer parameters:

[0007]

[0008] Where: Q c is the heat exchange capacity of the condenser, kJ·s -1 ;D c is the exhaust load of the low-pressure cylinder steam turbine, kg·s -1 ;h c is the exhaust steam specific enthalpy, kJ·kg -1 ;h c ' is the saturated steam enthalpy corresponding to the condenser pressure, kJ·kg -1 ; A c is the effective cooling area of ​​the condenser, m 2 ;t ci ,t co ,ts Respectively represent the condenser inlet water temperature, outlet water temperature, and condensate temperature, ℃; ρ w is cooling water density, kg·m -3 ;c w is the specific heat capacity of cooling water, kJ·kg -1 ℃ -1 ;P c is the condenser pressure, Pa; a is the saturated steam enthalpy function coefficient matrix:

[0009]

[0010] The iterative method is used to calculate the heat transfer process of the water pipeline. The iterative solution formula is as follows:

[0011]

[0012] Where: q smp is the heat dissipation per unit length of the pipe, J·s -1 ρ w is cooling water density, kg·m -3 ; W sys is the system cooling water flow, m 3 ·s -1 ;d po ,d pi are the outer diameter and inner diameter of the tube, m; t mpi ,t mpo are the inlet and outlet water temperatures of the main pipeline, ℃; t smpi ,t smpo are the inlet and outlet water temperatures of the sub-pipe section, ℃; h w is the convection heat transfer coefficient between the medium in the tube and the inner wall of the tube, W·m -2 ℃ -1 ; pt is the thermal conductivity of the tube, W·m -1 ℃ -1 ;h a is the convective heat transfer coefficient between ambient air and the outer wall of the tube, W·m -2 ℃ -1 ;

[0013] The outlet water temperature of the natural ventilation cooling tower of the power plant is obtained by iteratively solving the thermodynamic balance equation and the dynamic balance equation. The enthalpy difference method based on the Merkel model is used to calculate the heat exchange balance of the cooling tower, as shown below:

[0014]

[0015] Where: N' is the characteristic number of the cooling tower; K a is the bulk density per unit volume of the water-spraying filler, kg·m -3 ·s -1; V is the packing volume, m 3 ; N is the cooling number; K e To consider the coefficient of heat dissipation of evaporating water, which is related to the outlet water temperature; t towi ,t towo are the inlet and outlet temperatures of cooling water in the cooling tower, ℃; h t " is the saturated air enthalpy corresponding to the water temperature, kJ·kg -1 ;

[0016] The dynamic balance equation of the cooling tower is calculated as follows:

[0017]

[0018] Where: F p is the natural ventilation cooling tower suction, Pa; F d is the resistance of the natural ventilation cooling tower, Pa; H e is the effective height of the cooling tower, m; g is the acceleration due to gravity, m 2 ·s -1 ρ ai , ao are the air densities entering and leaving the tower, kg·m -3 ; ξ is the total resistance coefficient of the cooling tower; the subscript "m" represents the average value; v a is the air velocity, m·s -1 ;

[0019] After the system circulating water and the make-up water are mixed, the mixed water temperature is determined by the following formula:

[0020] t mix =[W ma t ma +(W c -W ma )t towo ] / W c

[0021] Where: t ma is the make-up water temperature, °C;

[0022] S2. Apply S1 to calculate and establish a system operating water temperature database, and use a deep learning method based on Bayesian optimization to establish a system water temperature black box model:

[0023]

[0024] Where: θ is the air temperature, ℃; is the relative humidity, %; D k is the condenser steam load, kg·s -1 ; W c is the cooling water flow rate of the condenser, m 3 ·s-1 ; K c is the condenser heat transfer coefficient, W·m -2 ℃ -1 ;

[0025] S3, establish the condenser tube net scaling rate model, and combine it with the water temperature black box model of S2 to correct the condenser operating water temperature;

[0026] The calculation formula for the net fouling rate of the condenser tube is as follows:

[0028] m s '=m d '-m r '

[0029] Where: m s ' is the net growth rate of dirt, kg·m -2 ·s -1 ;m d ' is the dirt deposition rate, kg·m -2 ·s -1 ;m r ' is the dirt peeling rate, kg·m -2 ·s -1 ;

[0030] After ΔT from time T, the total mass of scale on the inner wall of the condenser tube is expressed as:

[0031] m s,T+ΔT =m s,T +m s 'ΔT

[0032] Where: m s is the total mass of scale on the inner wall of the condenser, kg;

[0033] The total thermal resistance of scale is:

[0034] R f,T+ΔT =(δ f,T +m s 'ΔT / ρ f ) / λ f

[0035] Where: f,T is the thickness of the scale layer at time T, m; ρ f is the scale density, kg·m - 3; λ f is the thermal conductivity of dirt, W·m -1 ·K -1 ;

[0036] In summary, the condenser heat transfer coefficient, condenser inlet and outlet water temperature, and condensate temperature at time T+ΔT are expressed as:

[0037]

[0038] Where: K clean is the heat transfer coefficient of the condenser under clean conditions, W·m -2 ℃ -1 ;

[0039] S4. Taking the condenser operating water temperature, pipeline flow rate, and number of operating pumps as constraint parameters, considering the operating and start-up and shutdown costs of the circulating cooling water system pump unit, the system make-up water consumption costs, and the condenser cleaning costs, a two-level optimization model for maximum net profit during the power plant operation period is established;

[0040] The optimization model is expressed as:

[0041]

[0042] Where: I Top,max is the maximum net profit of the power plant during operation, yuan; I Tadj,max is the maximum net profit of the power plant in the jth circulating cooling water system regulation cycle, yuan; I total,j is the total revenue of the power plant steam turbine unit in the jth regulation cycle, RMB; C p,j is the operating cost of the pump unit of the circulating cooling water system in the jth regulation cycle, RMB; C ma,j is the water supplement cost for the jth regulation cycle, RMB; C ad,j is the pump unit regulation cost in the jth regulation cycle, which refers to the unit start-up and shutdown costs, RMB; C clean,j is the cleaning cost of the condenser rubber ball online cleaning device in the jth regulation cycle, yuan; δ f is the thickness of the scale layer on the condenser tube, m; n cs is the number of constant speed pumps in operation; n vf The number of variable frequency pumps in operation;

[0043] The total constraints of the optimization model are:

[0044]

[0045] Where: subscripts min and max represent the minimum and maximum values ​​respectively; W dt , W tco , W vc , W vm The corresponding cooling water flow rate is solved by reversely solving the condenser end difference, outlet water temperature, condenser pipe flow rate, and circulating cooling water system main line flow rate, m 3 ·s -1 ; W cd is the condenser design flow rate, m 3 ·s -1 ;n csi is the number of fixed speed pumps installed; nvfi I is the number of variable speed pumps installed; + represents the set of positive integers;

[0046] S5, solving the two-layer optimization model in S4, obtaining the optimal division scheme of the regulation period of the circulating cooling water system of the power plant, the optimal cleaning scheme of the condenser, and the optimal operation scheme of the pump unit in each regulation period;

[0047] The optimization model is solved using the atomic search algorithm. The algorithm calculation formula and improvement strategy are introduced as follows:

[0048] The atomic velocity and position update formulas are as follows:

[0049]

[0050] Where: v is the atomic velocity; subscripts i and j represent the i-th and j-th atoms; superscript d represents the atom for the d-th element in the variable vector; D is the vector dimension; k is the current iteration number; a is the atomic acceleration, which is calculated by the following formula:

[0051]

[0052] Where: F is the sum of the forces of the surrounding atoms on the current atom i, based on the Leonard Jones potential equation; α is the depth weight; β c is the coefficient factor; G is the constraint force of the current atom on the best atom; h is a parameter related to the distance between the two atoms;

[0053] The basic ASO algorithm is improved in the following steps: according to the results of multiple runs of the basic ASO, the fluctuation range R of the solution is determined; by rewriting the basic program, the solution corresponding to the best atom of each iteration is perturbed multiple times within the range set by R, and a new solution is determined to avoid the local optimum;

[0054] The optimal regulation cycle division scheme, the best condenser cleaning scheme and the optimal operation scheme of each regulation cycle pump unit in the operation period of the circulating cooling water system are solved; the optimal operation scheme of each regulation cycle pump unit includes: the number of fixed speed pump units in operation and the number of variable speed pump units in operation in each cycle and their motor frequency ratio;

[0055] S6. Using the climate parameters of the power plant in historical years, all the operating periods of the power plant throughout the year are connected in series according to the results of S4 and S5 to establish a model for the maximum net profit of the power plant throughout the year, which is used to calculate and evaluate the optimization potential of the circulating cooling water system of the power plant throughout the year;

[0056] The annual net income of the power plant is calculated as follows:

[0057] I year =I Top1,max +...+I Topl,max(K c,op(l-1) ,δ f,op(l-1) )+...+I Topz,max (K c,op(z-1) ,δ f,op(z-1) )

[0058] Where: I year is the annual net income of the power plant, yuan; I Topl,max is the maximum net profit of the power plant in the first operating period, yuan; K c,op(l-1) is the condenser heat transfer coefficient at the end of the l-1 operation period, W / (m 2 K);δ f,op(l-1) The thickness of the scale layer on the condenser tube at the end of the l-1 operation period, m; z is the total number of operation periods throughout the year;

[0059] When evaluating the potential for optimizing the operation of the circulating cooling water system throughout the year, the water saving and the increase in net profit of the power plant after system optimization are calculated based on the original system operation plan and the optimized operation plan.

[0060] Furthermore, the S2 Bayesian optimization process is as follows:

[0061] S2-1, discretize the input parameters within the specified range, use the method of S1 to calculate, obtain the condenser inlet water temperature, condenser outlet water temperature, and condensate water temperature corresponding to each set of environmental conditions and circulating cooling water system operation scheme, and establish a system operation water temperature database;

[0062] S2-2. Determine the three deep learning hyperparameters of Bayesian optimization: network depth, initial learning rate, and gradient descent momentum. The change of network depth is achieved by adjusting the number of convolution modules. The form of convolution modules is set as: convolution layer-normalization layer-activation layer.

[0063] S2-3. Deep learning completes the training of all sample data, and finally uses the validation set error as an indicator to judge the training effect. The absolute value of the difference between the water temperature prediction result and the actual value does not exceed 0.1℃ as the prediction qualification standard. The Bayesian optimization process takes the training set, validation set data, and optimized hyperparameters as input, and the prediction error of the convolutional neural network training result and the corresponding network structure as output. The optimization process is implemented with the minimum error rate of the test set as the goal. The Bayesian optimization objective function is expressed as:

[0064]

[0065] In the formula: num represents the number of qualified prediction standards; Y pre , Y val Respectively represent the predicted value and the actual value of the validation set; length represents the length operation of the matrix;

[0066] S2-4. Finally, the black box model is obtained.

[0067] Furthermore, the fouling deposition rate model and the peeling rate model in the net fouling rate model of the condenser tube in S3 are as follows:

[0068] According to Hasson's classical theory, with concentration difference as the driving force, the dirt deposition rate is expressed as:

[0069]

[0070] Where: β is the mass transfer coefficient from the main solution to the scale layer boundary, m·s -1 ; C F is the salt concentration in the flowing medium, kg·m -3 ; C f is the salt concentration at the interface between the fluid solution and the dirt, kg·m -3 ;k R is the reaction rate coefficient, m 4 kg -1 ·s -1 ; C s is the saturation concentration of sparingly soluble salts in the fluid, kg·m -3 ; n is the scaling reaction order. For reactions only related to anions and cations, n = 2;

[0071] Rearrange the above formula to

[0072]

[0073] Where: d ti is the inner diameter of the pipe, m; C F,Me , C F,An are respectively the concentrations of scaling anions and cations, kg·m -3 ; u is the cooling water flow rate in the tube, m·s -1 ; ν is the kinematic viscosity of water, m 2 ·s -1 ; The subscript ref indicates reference, t is the circulating water temperature in the condenser, K, which is the average temperature of the condenser inlet and outlet water; K R0 is the reaction rate constant, m 4 kg -1 ·s -1 ; E is the activation energy of the reaction, J·mol -1 ; R is the molar gas constant, which is 8.314 J·mol -1 ·K -1 , t d is the surface temperature of the scale layer, K;

[0074] Assuming that the temperature gradient of the scale layer is 0, the scale peeling rate equation is expressed as:

[0075] m r '=83.2u 2.54 ρ f d c (ρ 3 vg) 1 / 3 δ f

[0076] Where: f is the scale density, kg·m - 3;d c is the scale layer crystal diameter, m.

[0077] Furthermore, the total power generation revenue and various costs in the two-layer optimization model built by S4 are as follows:

[0078] The derivation process of the total power generation revenue formula of the power plant is as follows:

[0079] The general expression of the power generation of the steam turbine unit is:

[0080]

[0081] Where: b 1 、b 2 is the power correction factor; R s is the steam load rate; N rtur is the rated power of the steam turbine, kW; α r is the proportion of exhaust resistance loss of steam turbine;

[0082] By integrating the above formula over time, we can get the total power generation revenue within the cycle, and use the rectangular method to simplify the integral formula and solve it:

[0083]

[0084] Where: T is the duration of the adjustment cycle, h; n tur is the number of running steam turbines, e is the electricity price, yuan·kW -1 ·h -1 ,For some peak load plants, the electricity sales prices are different at different time periods;

[0085] The operating cost of the pump unit of the circulating cooling water system in period T is:

[0086]

[0087] Where: N p is the total operating power of the circulating cooling water system pump unit, kW, calculated by the following formula:

[0088]

[0089] Where: the subscripts "cs", "vf", and "ma" represent fixed speed pumps, variable frequency pumps, and water supply pumps, respectively; n is the number of pump units in operation; H is the pump head, m; η p , η tm , η m , η f are the efficiencies of the water pump, transmission device, motor, and inverter respectively;

[0090] The formula for calculating the supplementary water cost is as follows:

[0091] C ma =3600TW ma r ma

[0092] Where: r ma The unit price of supplementary water, yuan·m -3 ;

[0093] The calculation formula for the pump adjustment loss cost is as follows:

[0094] C ad =(|Δn vf |+|Δn cs |)p ad

[0095] Where: Δn is the difference between the number of running pumps in this regulation cycle and the number of running pumps in the previous regulation cycle; p ad The start-up and shutdown cost of a single water pump unit, RMB;

[0096] For power plants equipped with a rubber ball online cleaning system, the cleaning cost is expressed as follows:

[0097]

[0098] Where: N rp 、n rp They are respectively the operating power and the number of operations of the rubber ball pump, T rp is the running time of the rubber ball pump, h; num clean is the number of cleanings in this cycle, p r This is the additional cost for a single cleaning of the rubber ball, RMB.

[0099] Beneficial effects: Compared with the prior art, the advantages of the present invention are: solving the problem of mismatch between the operation of the circulating cooling water system of a power plant and the peak-shaving operation of the thermal system, and providing a scheme for optimizing and regulating the operation of the circulating cooling water system within a time period, including optimization of the regulation cycle division scheme within the time period and optimization of the operation scheme of each cycle, so as to maximize the net profit of the power plant during the operation period and achieve the purpose of energy saving and gain. BRIEF DESCRIPTION OF THE DRAWINGS

[0100] Figure 1The present invention provides a flow chart for determining the optimization scheme of the circulating cooling water system based on the maximum net profit of the power plant and calculating the system's annual optimization potential;

[0101] Figure 2 This is a diagram of a circulating cooling water system of a power plant according to an embodiment of the present invention;

[0102] Figure 3 This is a graph showing changes in environmental parameters during a certain operating period of a power plant according to an embodiment of the present invention;

[0103] Figure 4 This is a diagram showing changes in environmental parameters throughout the year in the area where the power plant is located according to an embodiment of the present invention;

[0104] Figure 5 The heat transfer calculation process of the circulating cooling water system according to the embodiment of the present invention;

[0105] Figure 6 The embodiment of the present invention adopts the Bayesian optimization method to optimize the iterative process of the deep learning network;

[0106] Figure 7 The deep learning network structure corresponding to the black box model of the optimal operating water temperature of the circulating cooling water system according to the embodiment of the present invention;

[0107] Figure 8 A diagram showing the number of times the operation scheme of the pump unit is adjusted in the operation cycle before and after the full-year operation optimization of the circulating cooling water system according to an embodiment of the present invention;

[0108] Fig. 9 The results of water saving and net benefit increase after the implementation of the present invention in optimizing the year-round operation of the circulating cooling water system of a power plant, wherein (a) is the water saving situation in each operating period throughout the year after optimization, (b) is the net profit of the power plant in each operating period throughout the year before and after optimization, and (c) is the added value of the net profit of the power plant in each operating period throughout the year after optimization. DETAILED DESCRIPTION

[0109] The following is a detailed description in conjunction with the specific implementation methods and the accompanying drawings.

[0110] like Figure 1 The cycle cooling water system operation optimization method based on the maximum net profit of the power plant shown includes the following steps:

[0111] S1. Establish heat transfer models for each component of the condenser circulating cooling water system of the power plant and perform comprehensive heat transfer calculations for the system;

[0112] S2, apply S1 to calculate and establish the system operation water temperature database, and use the deep learning method based on Bayesian optimization to establish the system water temperature black box model;

[0113] S3, establish the condenser tube net scaling rate model, and combine it with the water temperature black box model of S2 to correct the calculated condenser operating water temperature;

[0114] S4. Taking the condenser operating water temperature, pipeline flow rate, and number of operating pumps as constraint parameters, considering the operating and start-up and shutdown costs of the circulating cooling water system pump unit, the system make-up water consumption costs, and the condenser cleaning costs, a two-level optimization model for maximum net profit during the power plant operation period is established;

[0115] S5, solving the two-layer optimization model in S4, obtaining the optimal division scheme of the regulation period of the circulating cooling water system of the power plant, the optimal cleaning scheme of the condenser, and the optimal operation scheme of the pump unit in each regulation period;

[0116] S6. Using the climate parameters of the power plant in historical years, all operating periods of the power plant throughout the year are connected in series according to the results of S4 and S5 to establish a model for the maximum net profit of the power plant throughout the year, which is used to calculate and evaluate the optimization potential of the power plant's circulating cooling water system throughout the year.

[0117] Among them, the establishment and calculation process of S1 is:

[0118] According to the condenser heat transfer balance equation, the following equations are listed to solve the condenser heat transfer parameters:

[0119]

[0120] Where: Q c is the heat exchange capacity of the condenser, kJ·s -1 ;D c is the exhaust load of the low-pressure cylinder steam turbine, kg·s -1 ;h c is the exhaust steam specific enthalpy, kJ·kg -1 ;h c ' is the saturated steam enthalpy corresponding to the condenser pressure, kJ·kg -1 ; A c is the effective cooling area of ​​the condenser, m 2 ;t ci ,t co ,t s Respectively represent the condenser inlet water temperature, outlet water temperature, and condensate temperature, ℃; ρ w is cooling water density, kg·m -3 ;c w is the specific heat capacity of cooling water, kJ·kg -1 ℃ -1 ;P c is the condenser pressure, Pa; a is the coefficient matrix of the saturated steam enthalpy function, as shown in formula (2):

[0121]

[0122] The iterative method is used to calculate the heat transfer process of the water pipeline. The iterative solution formula is as follows:

[0123]

[0124] Where: q smp is the heat dissipation per unit length of the pipe, J·s -1 ρ w is cooling water density, kg·m -3 ; W sys is the system cooling water flow, m 3 ·s -1 ;d po d pi are the outer diameter and inner diameter of the tube, m; t mpi ,t mpo are the inlet and outlet water temperatures of the main pipeline, ℃; t smpi ,t smpo are the inlet and outlet water temperatures of the sub-pipe section, ℃; h w is the convection heat transfer coefficient between the medium in the tube and the inner wall of the tube, W·m -2 ℃ -1 ; pt is the thermal conductivity of the tube, W·m -1 ℃ -1 ;h a is the convective heat transfer coefficient between ambient air and the outer wall of the tube, W·m -2 ℃ -1 ;

[0125] The outlet water temperature of the natural ventilation cooling tower of the power plant is obtained by iteratively solving the thermodynamic balance equation and the dynamic balance equation. The enthalpy difference method based on the Merkel model is used to calculate the heat exchange balance of the cooling tower, as shown below:

[0126]

[0127] Where: N' is the characteristic number of the cooling tower; K a is the bulk density per unit volume of the water-spraying filler, kg·m -3 ·s -1 ; V is the packing volume, m 3 ; N is the cooling number; K e To consider the coefficient of heat dissipation of evaporating water, which is related to the outlet water temperature; t towi ,t towo are the inlet and outlet temperatures of cooling water in the cooling tower, ℃; h t " is the saturated air enthalpy corresponding to the water temperature, kJ·kg -1 ;

[0128] The dynamic balance equation of the cooling tower is calculated as follows:

[0129]

[0130] Where: F p is the natural ventilation cooling tower suction, Pa; F d is the resistance of the natural ventilation cooling tower, Pa; H e is the effective height of the cooling tower, m; g is the acceleration due to gravity, m 2 ·s -1 ρ ai , ao are the air densities entering and leaving the tower, kg·m -3 ; ξ is the total resistance coefficient of the cooling tower; the subscript "m" represents the average value; v a is the air velocity, m·s -1 ;

[0131] After the system circulating water and the make-up water are mixed, the mixed water temperature is determined by the following formula:

[0132] t mix =[W ma t ma +(W c -W ma )t towo ] / W c (6)

[0133] Where: t ma is the make-up water temperature, ℃.

[0134] S2 specifically includes:

[0135] S2-1, discretize the input parameters within the specified range, use the method of S1 to calculate, obtain the condenser inlet water temperature, condenser outlet water temperature, and condensate water temperature corresponding to each set of environmental conditions and circulating cooling water system operation scheme, and establish a system operation water temperature database;

[0136] S2-2. Determine the three deep learning hyperparameters of Bayesian optimization: network depth, initial learning rate, and gradient descent momentum. The change of network depth is achieved by adjusting the number of convolution modules. The form of convolution modules is set as: convolution layer-normalization layer-activation layer.

[0137] S2-3. Deep learning completes the training of all sample data, and finally uses the validation set error as an indicator to judge the training effect. The absolute value of the difference between the water temperature prediction result and the actual value does not exceed 0.1℃ as the prediction qualification standard. The Bayesian optimization process takes the training set, validation set data, and optimized hyperparameters as input, and the prediction error of the convolutional neural network training result and the corresponding network structure as output. The optimization process is implemented with the minimum error rate of the test set as the goal. The Bayesian optimization objective function is expressed as:

[0138]

[0139] In the formula: num represents the number of qualified prediction standards; Y pre , Y val Respectively represent the predicted value and the actual value of the validation set; length represents the length operation of the matrix;

[0140] S2-4. The final black box model is expressed as:

[0141]

[0142] Where: θ is the air temperature, ℃; is relative humidity, %; D k is the condenser steam load, kg·s -1 ; W c is the cooling water flow rate of the condenser, m 3 ·s -1 ; K c is the condenser heat transfer coefficient, W·m -2 ℃ -1 .

[0143] S3 is specifically:

[0144] The calculation formula for the net fouling rate of the condenser tube is as follows:

[0146] m s '=m d '-m r ' (9) Where: m s ' is the net growth rate of dirt, kg·m -2 ·s -1 ;m d ' is the dirt deposition rate, kg·m -2 ·s -1 ;m r ' is the dirt peeling rate, kg·m -2 ·s -1 ;

[0147] According to Hasson's classic theory: on the basis of the existing first layer of scaling, ions adhere to the surface of the scaling layer through the flow of the main solution, and then enter the lattice composed of atoms, molecules or ions to form insoluble salt crystals; with the concentration difference as the driving force, the scaling deposition rate is expressed as:

[0148]

[0149] Where: β is the mass transfer coefficient from the main solution to the scale layer boundary, m·s -1 ; CF is the salt concentration in the flowing medium, kg·m -3 ; C f is the salt concentration at the interface between the fluid solution and the dirt, kg·m -3 ;k R is the reaction rate coefficient, m 4 kg -1 ·s -1 ; C s is the saturation concentration of sparingly soluble salts in the fluid, kg·m -3 ; n is the scaling reaction order. For reactions only related to anions and cations, n = 2;

[0150] Rearrange the above formula to

[0151]

[0152] Where: d ti is the inner diameter of the pipe, m; C F,Me , C F,An are respectively the concentrations of scaling anions and cations, kg·m -3 ; u is the cooling water flow rate in the tube, m·s -1 ; ν is the kinematic viscosity of water, m 2 ·s -1 ; The subscript ref indicates reference, t is the circulating water temperature in the condenser, K, which is the average temperature of the condenser inlet and outlet water; K R0 is the reaction rate constant, m 4 kg -1 ·s -1 ; E is the activation energy of the reaction, J·mol -1 ; R is the molar gas constant, which is 8.314 J·mol -1 ·K -1 , t d is the surface temperature of the scale layer, K;

[0153] When the scale grows to a certain thickness, it will peel off due to the shear force. Assuming that the temperature gradient of the scale layer is 0, the scale peeling rate equation is simplified to:

[0154] m r '=83.2u 2.54 ρ f d c (ρ 3 vg) 1 / 3 δ f (12)

[0155] Where: f is the scale density, kg·m - 3;d c is the scale crystal diameter, m;

[0156] Through the progressive calculation of the time step within the cycle, the quantitative analysis of the evolution of the condenser tube scale thickness with the time step is realized, and the condenser heat transfer coefficient, condenser inlet and outlet water temperature, and condensate temperature at any time are determined; for the time period [T, T+ΔT], the initial value of the condenser heat transfer coefficient K is given c,T and the initial value of the scale thickness δ f,T , according to the black box model, the condenser inlet and outlet water temperatures t at the initial moment are calculated ci,T ,t co,T ; After ΔT from time T, the total mass of scale on the inner wall of the condenser tube is expressed as:

[0157] m s,T+ΔT =m s,T +m s 'ΔT (13)

[0158] The total thermal resistance of scale is:

[0159] R f,T+ΔT =(δ f,T +m s 'ΔT / ρ f ) / λ f (14)

[0160] Where: f,T is the thickness of the scale layer at time T, m; λ f is the thermal conductivity of dirt, W·m -1 ·K -1 ;

[0161] In summary, the condenser heat transfer coefficient, condenser inlet and outlet water temperature, and condensate temperature at time T+ΔT are expressed as:

[0162]

[0163] Where: K clean is the heat transfer coefficient of the condenser under clean conditions, W·m -2 ℃ -1 .

[0164] S4 is specifically:

[0165] The inner layer of the model aims to maximize the net profit of the power plant during the regulation period of the circulating cooling water system, and optimizes the number of fixed-speed pump units in operation, the number of variable-speed pump units in operation and their frequency conversion ratio, and the condenser cleaning plan; the outer layer aims to maximize the net profit of the power plant during the operation period, and optimizes the specified ambient temperature difference used to determine the regulation period division plan. It should be pointed out that each regulation cycle must be optimized in series, because the results of the operation calculation part of the previous cycle will be used as the initial value of the next cycle, and the associated parameters of two adjacent cycles have been noted in formula (16).

[0166] The optimization model is expressed as:

[0167]

[0168] Where: I Top,max is the maximum net profit of the power plant during operation, yuan; I Tadj,max is the maximum net profit of the power plant in the jth circulating cooling water system regulation cycle, yuan; I total,j is the total revenue of the power plant steam turbine unit in the jth regulation cycle, RMB; C p,j is the operating cost of the pump unit of the circulating cooling water system in the jth regulation cycle, RMB; C ma,j is the water supplement cost for the jth regulation cycle, RMB; C ad,j is the pump unit regulation cost in the jth regulation cycle, mainly referring to the unit start-up and shutdown costs, RMB; C clean,j is the cleaning cost of the condenser rubber ball online cleaning device in the jth regulation cycle, yuan; δ f is the thickness of the scale layer on the condenser tube, m; n cs is the number of constant speed pumps in operation; n vf The number of variable frequency pumps in operation;

[0169] The total constraints of the optimization model are:

[0170]

[0171] Where: subscripts min and max represent the minimum and maximum values ​​respectively; W dt , W tco , W vc , W vm The corresponding cooling water flow rate is solved by reversely solving the condenser end difference, outlet water temperature, condenser pipe flow rate, and circulating cooling water system main line flow rate, m 3 ·s -1 ; W cd is the condenser design flow rate, m 3 ·s -1 ;n csi is the number of fixed speed pumps installed; n vfi I is the number of variable speed pumps installed; + represents the set of positive integers;

[0172] The derivation process of the total power generation revenue formula of the power plant is as follows:

[0173] The general expression of the power generation of the steam turbine unit is:

[0174]

[0175] Where: b 1 、b 2 is the power correction factor; R s is the steam load rate; Nrtur is the rated power of the steam turbine, kW; α r is the proportion of exhaust resistance loss of steam turbine;

[0176] By integrating equation (18) over time, the total power generation revenue within the cycle is obtained, as shown in the middle of equation (19). The rectangular method is used to simplify the integral equation, as shown in the formula on the right side of equation (19):

[0177]

[0178] Where: T is the duration of the adjustment cycle, h; n tur is the number of running steam turbines, e is the electricity price, yuan·kW -1 ·h -1 ,For some peak load plants, the electricity sales prices are different at different time periods;

[0179] The operating cost of the pump unit of the circulating cooling water system within the cycle is:

[0180]

[0181] Where: N p is the total operating power of the circulating cooling water system pump unit, kW, calculated by the following formula:

[0182]

[0183] Where: the subscripts "cs", "vf", and "ma" represent fixed speed pumps, variable frequency pumps, and water supply pumps, respectively; n is the number of pump units in operation; H is the pump head, m; η p , η tm , η m , η f are the efficiencies of the water pump, transmission device, motor, and inverter respectively;

[0184] The formula for calculating the supplementary water cost is as follows:

[0185] C ma =3600TW ma r ma (twenty two)

[0186] Where: r ma The unit price of supplementary water, yuan·m -3 ;

[0187] The calculation formula for the pump adjustment loss cost is as follows:

[0188] C ad =(|Δn vf |+|Δn cs |)p ad (twenty three)

[0189] Where: Δn is the difference between the number of running pumps in this regulation cycle and the number of running pumps in the previous regulation cycle; p ad The start-up and shutdown cost of a single water pump unit, RMB;

[0190] For power plants equipped with a rubber ball online cleaning system, the condenser cleaning time and frequency can be flexibly adjusted. The cleaning cost is expressed as follows:

[0191]

[0192] Where: N rp 、n rp They are respectively the operating power and the number of operations of the rubber ball pump, T rp is the running time of the rubber ball pump, h; num clean is the number of cleanings in this cycle, p r This is the additional cost for a single cleaning of the rubber ball, RMB.

[0193] In S5:

[0194] The optimization model is solved using the atomic search algorithm. The algorithm calculation formula and improvement strategy are introduced as follows:

[0195] The atomic velocity and position update formulas are as follows:

[0196]

[0197] Where: v is the atomic velocity; subscripts i and j represent the i-th and j-th atoms; superscript d represents the atom for the d-th element in the variable vector; D is the vector dimension; k is the current iteration number; a is the atomic acceleration, which is calculated by the following formula:

[0198]

[0199] Where: F is the sum of the forces of the surrounding atoms on the current atom i, based on the Leonard Jones potential equation; α is the depth weight; β c is the coefficient factor; G is the constraint force of the current atom on the best atom; h is a parameter related to the distance between the two atoms;

[0200] In order to improve the convergence performance of the algorithm, the basic ASO algorithm is improved. The steps are as follows: according to the results of multiple runs of the basic ASO, the fluctuation range R of the solution is determined; by rewriting the basic program, the solution corresponding to the best atom of each iteration is perturbed multiple times within the range set by R, and a new solution is determined to avoid local optimality;

[0201] The optimal regulation cycle of the operating plan of the circulating cooling water system during the operating period and the optimal operating plan of the pump units in each regulation cycle are solved; the optimal operating plan of the pump units in each regulation cycle includes: the number of fixed-speed pump units in operation in each cycle, the number of variable-speed pump units in operation, and the motor frequency ratio.

[0202] In S6:

[0203] The annual net income of the power plant is calculated as follows:

[0204] I year =I Top1,max +...+I Topl,max (K c,op(l-1) ,δ f,op(l-1) )+...+I Topz,max (K c,op(z-1) ,δ f,op(z-1) )(27)

[0205] Where: I year is the annual net income of the power plant, yuan; I Topl,max is the maximum net profit of the power plant in the lth operating period, yuan; K c,op(l-1) is the condenser heat transfer coefficient at the end of the l-1 operation period, W / (m 2 K);δ f,op(l-1) The thickness of the scale layer on the condenser tube at the end of the l-1 operation period, m; z is the total number of operation periods throughout the year;

[0206] When evaluating the potential for optimizing the operation of the circulating cooling water system throughout the year, the water saving and the increase in net profit of the power plant after system optimization are calculated based on the original system operation plan and the optimized operation plan.

[0207] The specific implementation mode of the present invention is further described below through a specific application case. The main technical parameters of the power plant are as follows:

[0208] A power plant is equipped with two 330MW steam turbine units, and the supporting circulating cooling water system is arranged as follows: Figure 2 As shown. The power plant is a peak-shaving power plant. Through grid dispatching, the power demand side can receive all the electricity generated by the power plant, mainly providing electricity for urban residents. For electricity that cannot be consumed in the urban area, the power plant will cooperate with other power plants to supply it to industrial enterprises around the urban area. During the peak electricity consumption period (8:00-21:00), the steam turbine unit operates at full load, and the on-grid electricity price is 0.4691 yuan / kWh. During the low electricity consumption period (0:00-8:00, 21:00-24:00), the steam load rate is 0.6 and the electricity price is 0.3153 yuan / kWh. The changes in environmental parameters in a certain operating period and the changes in environmental parameters throughout the year in the area where the power plant is located are respectively Figure 3 and Figure 4 The circulating cooling water system is regulated by adjusting the number of pump units in operation on a quarterly basis, with 3 units started in spring and autumn, 4 units in summer, and 2 units in winter. The system is equipped with 3 fixed-speed pump units of the same model and 1 variable-frequency pump unit. The pump performance equation is shown below.

[0209]

[0210] Where: H is the pump head, m; Q is the pump flow, m 3 ·s -1 ; η is the pump efficiency, %.

[0211] The model of the power plant's steam turbine is N330-16.67 / 538 / 538, and its main technical parameters are shown in Table 1.

[0212] Table 1 Main design parameters of steam turbine

[0213]

[0214] The model of the condenser of the circulating cooling water system is N-20248, and its main technical parameters are shown in Table 2.

[0215] Table 2 Main design parameters of condenser

[0216]

[0217] Step A: Establish the heat transfer model of each component of the power plant condenser circulating cooling water system and perform comprehensive heat transfer calculation of the system.

[0218] Figure 5 The ambient dry bulb temperature is 19°C, the relative humidity is 80%, and the condenser steam flow rate is 175 kg·s -1 , condenser cooling water flow rate is 7.5m 3 ·s -1 The iterative calculation process of the system operating water temperature under working conditions. As the number of iterations increases, the inlet and outlet water temperatures of each link rise first, then reach a stable value at the 6th iteration, and then remain unchanged. The water temperature at this time is finally solved as the stable operating water temperature.

[0219] Step B: Establish an operating water temperature database based on the circulating heat transfer calculation results of the circulating cooling water system, and use a deep learning method based on Bayesian optimization to establish a water temperature black box model.

[0220] The calculation results of circulating heat transfer in the circulating cooling water system are used as sample data, and these samples are trained through deep learning to achieve nonlinear regression prediction with 5 inputs and 3 outputs - steam load, condenser heat transfer coefficient, dry bulb temperature, relative humidity, condenser cooling water flow rate as input, condenser inlet and outlet water temperature, condensate temperature as output. Example The sample size of the water temperature database of the circulating cooling water system of a power plant is about 200,000.

[0221] With the goal of minimizing the error of the network test set, the Bayesian optimization method was used to optimize the momentum coefficient, initial learning rate, and network depth of the multi-layer convolutional neural network. The Bayesian optimization process is as follows: Figure 6As shown. After optimization, the error of the model validation set reached 0.0053, which means that the difference between the output variables of 99.47% of the validation samples and the absolute value of the theoretical calculated water temperature is less than 0.1℃, which fully meets the calculation accuracy requirements. The optimal network hyperparameters obtained by optimization are set as follows: momentum drop coefficient is 0.9776, initial learning rate: 5.9875e-05, number of network layers: 13 layers, and the network structure is as follows Figure 7 The number of convolutional modules is a key parameter to control the depth of the network. Each module is composed of a convolutional layer, a normalization layer, and an activation layer, which are marked as Figure 7 The dashed box in .

[0222] Step C: Establish a condenser tube net scaling rate model, and combine it with the water temperature black box model of step B to correct the condenser operating water temperature.

[0223] The calculation process is corrected with a dry bulb temperature of 35°C, a relative humidity of 80%, and a steam load rate of 0.6. The operation plan of the circulating cooling water system is set as follows: 3 fixed-speed pump units are in operation, 1 variable-speed pump unit is in operation, and the frequency conversion ratio is 0.85. The calculation time is selected as 20h, and the initial heat transfer coefficient of the condenser is 3068.7W·m -2 ℃ -1 The initial scale thickness of the condenser tube is 1×10 -10 m.

[0224] Substitute the above operating parameters into the water temperature black box model:

[0225]

[0226] The condenser inlet water temperature, outlet water temperature, and condensed water temperature at the initial moment are 35.1802°C, 40.4624°C, and 42.0662°C, respectively. Then, the calculated water temperature is substituted into the condenser tube deposition rate calculation formula:

[0227]

[0228] Where: β is the mass transfer coefficient from the main solution to the scale layer boundary, m·s -1 ; C F is the salt concentration in the flowing medium, kg·m -3 ; C f is the salt concentration at the interface between the fluid solution and the dirt, kg·m -3 ;k R is the reaction rate coefficient, m 4 kg -1 ·s -1 ; C s is the saturation concentration of sparingly soluble salts in the fluid, kg·m -3 ; n is the scaling reaction order. For reactions only related to anions and cations, n = 2; dti is the inner diameter of the pipe, m; C F,Me , C F,An are respectively the concentrations of scaling anions and cations, kg·m -3 , in the embodiment, they are 0.04 kg·m -3 and 0.12kg·m -3 ; u is the cooling water flow rate in the tube, m·s -1 ; ν is the kinematic viscosity of water, m 2 ·s -1 ; The subscript ref indicates reference; t is the circulating water temperature in the condenser tube, K, which is the average temperature of the inlet and outlet water of the condenser; K R0 is the reaction rate constant, m 4 kg -1 ·s -1 , the example takes 1.69×10 21 m 4 kg -1 ·s -1 ; E is the activation energy of the reaction, J·mol -1 , the example takes 1.70×10 5 J.mol -1 ; R is the molar gas constant, which is 8.314 J·mol -1 ·K -1 , t d is the surface temperature of the scale layer, K.

[0229] Solving the above equation, we can obtain the initial deposition rate of the condenser tube as 5.6599×10 -9 kg·m -2 ·s -1 .

[0230] The condenser water temperature calculated at the initial moment is brought into the condenser tube scale peeling rate model:

[0231] m r '=83.2u 2.54 ρ f d c (ρ 3 vg) 1 / 3 δ f

[0232] Where: f is the scale density, kg·m - 3, Example 2.7×10 3 kg·m -3 ;d c is the diameter of the scale crystal, m, and in the embodiment, 3.6×10 -6 m.

[0233] The calculated scaling rate at the initial moment is 7.1441×10 -15 kg·m -2 ·s -1 , so far, the initial condensation net fouling rate can be calculated: m s '=m d '-m r '=5.6599×10 -9 -7.1441×10 -15 ≈5.6599×10 - 9 kg·m -2 ·s -1 The total mass of scale on the inner wall of the condenser tube, the total thermal resistance of scale, and the final corrected water temperature are calculated by equations (17) to (19).

[0234] After iterative calculation according to the above scheme to the last moment (running for 20 hours), the total thermal resistance of scale R f,T+ΔT is 3.663×10 -6 , the condenser heat transfer coefficient is: K c,T+ΔT =1 / (R f,T+ΔT +1 / K clean )=1 / (3.663×10 -6 +1 / 3068.7)=3034.5W·m -2 ℃ -1 The condenser operating water temperature correction calculation is:

[0235]

[0236] The final calculated condenser inlet water temperature, outlet water temperature and condensed water temperature are 35.1875℃, 40.4693℃ and 42.0977℃ respectively.

[0237] Step D: Establish a two-layer optimization model for the circulating cooling water system of the power plant:

[0238]

[0239] Where: I Top,max is the maximum net profit of the power plant during operation, yuan; I Tadj,max is the maximum net profit of the power plant in the jth circulating cooling water system regulation cycle, yuan; I total,j is the total revenue of the power plant steam turbine unit in the jth regulation cycle, RMB; C p,j is the operating cost of the pump unit of the circulating cooling water system in the jth regulation cycle, RMB; C ma,j is the water supplement cost for the jth regulation cycle, RMB; C ad,j is the pump unit regulation cost in the jth regulation cycle, mainly referring to the unit start-up and shutdown costs, RMB; C clean,jis the cleaning cost of the condenser rubber ball online cleaning device in the jth regulation cycle, yuan; δ f is the thickness of the scale layer on the condenser tube, m; n cs is the number of constant speed pumps in operation; n vf The number of variable frequency pumps in operation.

[0240] The overall constraints and parameter settings of the optimization model are as follows:

[0241]

[0242] Where: subscripts min and max represent the minimum and maximum values ​​respectively; W dt , W tco , W vc , W vm The corresponding cooling water flow rate is solved by reversely solving the condenser end difference, outlet water temperature, condenser pipe flow rate, and circulating cooling water system main line flow rate, m 3 ·s -1 In the embodiment, these values ​​are 10.11m 3 ·s -1 、11.22m 3 ·s -1 、9.98m 3 ·s -1 、11.57m 3 ·s -1 ; W cd is the condenser design flow rate, m 3 ·s -1 In the embodiment, the value is 10m 3 ·s -1 ;n csi is the number of fixed speed pumps installed, and in the embodiment, the value is 3; n vfi is the number of variable speed pumps installed, and in the embodiment, the value is 1; + Represents the set of positive integers.

[0243] Step E: Using the improved atomic search algorithm to solve the optimization model, the environmental parameters of the power plant during a certain operation period in the embodiment are as follows: Figure 3 The operation period is 48 hours, with a total of 144 time steps. The optimal operation plan of the pump unit obtained by optimization is shown in Table 3. The condenser cleaning is performed once, when the system runs to 2600 minutes. The total net profit of the power plant during the operation period obtained by optimization calculation is 10.925 million yuan.

[0244] Table 3 Optimization results of pump unit operation plan during a certain period of operation in the power plant

[0245]

[0246] Step F: According to formula (27), the operation optimization calculation of the circulating cooling water system of the power plant is carried out throughout the year. The annual environmental parameters are as follows: Figure 4 As shown. Figure 8 As shown, before optimization, the operation plan of the circulating cooling water system was only adjusted on a quarterly basis, without considering the changes in steam cooling load and environmental parameters during peak load regulation of the power plant; after optimization, the number of pump unit adjustments in each operating period increased significantly, at least 3 times. In each operating period of the embodiment power plant, there are 2 low-load periods and 2 high-load periods. Each peak load regulating circulating cooling water system should adjust the cooling water flow to match the steam condensation cooling load after peak load regulation. On this basis, the adjustment cycle is divided according to the change law of the ambient dry-bulb temperature in each equal load period. It can be seen that the adjustment of the power plant operation plan after optimization is more reasonable than before optimization.

[0247] The optimization method of the present invention enables power plants to save water and significantly improve economic benefits. Fig. 9 When the CCWS of the power plant was not optimized, it would operate in a large cooling water flow mode for a long time, requiring a large amount of make-up water. After the optimization scheme, the cooling water flow rate in each regulation cycle was reduced, and the annual cooling make-up water volume was reduced by 31.48×10 5 m 3 ( Fig. 9 (a)). Compared with the system without operation optimization, the optimized power plant's annual net profit increased by 4.5 million yuan (9(b), (c)), which is a significant achievement.

Claims

1. A method for optimizing the period operation of a circulating cooling water system based on the maximum net profit of a power plant, characterized in that: The following steps are involved: S1. Establish heat transfer models for each component of the condenser circulating cooling water system of the power plant and perform comprehensive heat transfer calculations for the system; According to the condenser heat transfer balance equation, the following equations are listed to solve the condenser heat transfer parameters: Where: Q c is the heat exchange capacity of the condenser, kJ·s -1 ;D c is the exhaust load of the low-pressure cylinder steam turbine, kg·s -1 ;h c is the exhaust steam specific enthalpy, kJ·kg -1 ;h c ' is the saturated steam enthalpy corresponding to the condenser pressure, kJ·kg -1 ; A c is the effective cooling area of ​​the condenser, m 2 ; t ci ,t co ,t s Respectively represent the condenser inlet water temperature, outlet water temperature, and condensate temperature, ℃; ρ w is cooling water density, kg·m -3 ;c w is the specific heat capacity of cooling water, kJ·kg -1 ℃ -1 ;P c is the condenser pressure, Pa; a is the coefficient matrix of the saturated steam enthalpy function, The iterative method is used to calculate the heat transfer process of the water pipeline. The iterative solution formula is as follows: Where: q smp is the heat dissipation per unit length of the pipe, J·s -1 ; ρ w is cooling water density, kg·m -3 ; W sys is the system cooling water flow, m 3 ·s -1 ;d po d pi are the outer diameter and inner diameter of the tube, m; t mpi ,t mpo are the inlet and outlet water temperatures of the main pipeline, ℃; t smpi ,t smpo are the inlet and outlet water temperatures of the sub-pipe section, ℃; h w is the convection heat transfer coefficient between the medium in the tube and the inner wall of the tube, W·m -2 ℃ -1 ; λ pt is the thermal conductivity of the tube, W·m -1 ℃ -1 ;h a is the convective heat transfer coefficient between ambient air and the outer wall of the tube, W·m -2 ℃ -1 ; The outlet water temperature of the natural ventilation cooling tower of the power plant is obtained by iteratively solving the thermodynamic balance equation and the dynamic balance equation. The enthalpy difference method based on the Merkel model is used to calculate the heat exchange balance of the cooling tower, as shown below: Where: N' is the characteristic number of the cooling tower; K a is the bulk density per unit volume of water-spraying filler, kg·m -3 ·s -1 ; V is the packing volume, m 3 ; N is the cooling number; K e To consider the coefficient of heat dissipation of evaporating water, which is related to the outlet water temperature; t towi ,t towo are the inlet and outlet temperatures of cooling water in the cooling tower, ℃; h t " is the saturated air enthalpy corresponding to the water temperature, kJ·kg -1 ; The dynamic balance equation of the cooling tower is calculated as follows: Where: F p is the natural ventilation cooling tower suction, Pa; F d is the resistance of the natural ventilation cooling tower, Pa; H e is the effective height of the cooling tower, m; g is the acceleration due to gravity, m 2 ·s -1 ρ ai , ao are the air densities entering and leaving the tower, kg·m -3 ; ξ is the total resistance coefficient of the cooling tower; the subscript "m" represents the average value; v a is the air velocity, m·s -1 ; After the system circulating water and the make-up water are mixed, the mixed water temperature is determined by the following formula: t mix =[W ma t ma +(W c -W ma )t towo ] / W c Where: t ma is the make-up water temperature, °C; S2. Apply S1 to calculate and establish a system operating water temperature database, and use a deep learning method based on Bayesian optimization to establish a system water temperature black box model: Where: θ is the air temperature, ℃; is relative humidity, %; D k is the condenser steam load, kg·s -1 ; W c is the cooling water flow rate of the condenser, m 3 ·s -1 ; K c is the condenser heat transfer coefficient, W·m -2 ℃ -1 ; S3, establish the condenser tube net scaling rate model, and combine it with the water temperature black box model of S2 to correct the condenser operating water temperature; The calculation formula for the net fouling rate of the condenser tube is as follows: m s '=m d '-m r ' Where: m s ' is the net growth rate of dirt, kg·m -2 ·s -1 ;m d ' is the dirt deposition rate, kg·m -2 ·s -1 ;m r ' is the dirt peeling rate, kg·m -2 ·s -1 ; After ΔT from time T, the total mass of scale on the inner wall of the condenser tube is expressed as: m s,T+ΔT =m s,T +m s 'ΔT Where: m s is the total mass of scale on the inner wall of the condenser, kg; The total thermal resistance of scale is: R f,T+ΔT =(δ f,T +m s 'ΔT / p f ) / min f Where: f,T is the thickness of the scale layer at time T, m; ρ f is the scale density, kg·m -3 ; f is the thermal conductivity of dirt, W·m -1 ·K -1 ; In summary, the condenser heat transfer coefficient, condenser inlet and outlet water temperature, and condensate temperature at time T+ΔT are expressed as: Where: K clean is the heat transfer coefficient of the condenser under clean conditions, W·m -2 ℃ -1 ; S4. Taking the condenser operating water temperature, pipeline flow rate, and number of operating pumps as constraint parameters, considering the operating and start-up and shutdown costs of the circulating cooling water system pump unit, the system make-up water consumption costs, and the condenser cleaning costs, a two-level optimization model for maximum net profit during the power plant operation period is established; The optimization model is expressed as: Where: I Top,max is the maximum net profit of the power plant during operation, yuan; I Tadj,max is the maximum net profit of the power plant in the jth circulating cooling water system regulation cycle, yuan; I total,j is the total revenue of the power plant steam turbine unit in the jth regulation cycle, RMB; C p,j is the operating cost of the pump unit of the circulating cooling water system in the jth regulation cycle, RMB; C ma,j is the water supplement cost for the jth regulation cycle, RMB; C ad,j is the pump unit regulation cost in the jth regulation cycle, which refers to the unit start-up and shutdown costs, RMB; C clean,j is the cleaning cost of the condenser rubber ball online cleaning device in the jth regulation cycle, yuan; δ f is the thickness of the scale layer on the condenser tube, m; n cs is the number of constant speed pumps in operation; n vf The number of variable frequency pumps in operation; The total constraints of the optimization model are: Where: subscripts min and max represent the minimum and maximum values ​​respectively; W dt , W tco , W vc , W vm The corresponding cooling water flow rate is solved by reversely solving the condenser end difference, outlet water temperature, condenser pipe flow rate, and circulating cooling water system main line flow rate, m 3 ·s -1 ; W cd is the condenser design flow rate, m 3 ·s -1 ;n csi is the number of fixed speed pumps installed; n vfi I is the number of variable speed pumps installed; + represents the set of positive integers; S5, solving the two-layer optimization model in S4, obtaining the optimal division scheme of the regulation period of the circulating cooling water system of the power plant, the optimal cleaning scheme of the condenser, and the optimal operation scheme of the pump unit in each regulation period; The optimization model is solved using the atomic search algorithm. The algorithm calculation formula and improvement strategy are introduced as follows: The atomic velocity and position update formulas are as follows: Where: v is the atomic velocity; subscripts i and j represent the i-th and j-th atoms; superscript d represents the atom for the d-th element in the variable vector; D is the vector dimension; k is the current iteration number; a is the atomic acceleration, which is calculated by the following formula: Where: F is the sum of the forces of the surrounding atoms on the current atom i, based on the Leonard Jones potential equation; α is the depth weight; β c is the coefficient factor; G is the constraint force of the current atom on the best atom; h is a parameter related to the distance between the two atoms; The basic ASO algorithm is improved in the following steps: according to the results of multiple runs of the basic ASO, the fluctuation range R of the solution is determined; by rewriting the basic program, the solution corresponding to the best atom of each iteration is perturbed multiple times within the range set by R, and a new solution is determined to avoid the local optimum; The optimal regulation cycle division scheme, the best condenser cleaning scheme and the optimal operation scheme of each regulation cycle pump unit in the operation period of the circulating cooling water system are solved; the optimal operation scheme of each regulation cycle pump unit includes: the number of fixed speed pump units in operation and the number of variable speed pump units in operation in each cycle and their motor frequency ratio; S6. Using the climate parameters of the power plant in historical years, all the operating periods of the power plant throughout the year are connected in series according to the results of S4 and S5 to establish a model for the maximum net profit of the power plant throughout the year, which is used to calculate and evaluate the optimization potential of the circulating cooling water system of the power plant throughout the year; The annual net income of the power plant is calculated as follows: I year =I Top1,max +...+I Topl,max (K c,op(l-1) ,δ f,op(l-1) )+...+I Topz,max (K c,op(z-1) ,δ f,op(z-1) ) Where: I year is the annual net income of the power plant, yuan; I Topl,max is the maximum net profit of the power plant in the first operating period, yuan; K c,op(l-1) is the condenser heat transfer coefficient at the end of the l-1 operation period, W / (m 2 K);δ f,op(l-1) The thickness of the scale layer on the condenser tube at the end of the l-1 operation period, m; z is the total number of operation periods throughout the year; When evaluating the potential for optimizing the operation of the circulating cooling water system throughout the year, the water saving and the increase in net profit of the power plant after system optimization are calculated based on the original system operation plan and the optimized operation plan.

2. The method for optimizing the period operation of a circulating cooling water system based on the maximum net profit of a power plant according to claim 1, characterized in that: The S2 Bayesian optimization process is as follows: S2-1, discretize the input parameters within the specified range, use the method of S1 to calculate, obtain the condenser inlet water temperature, condenser outlet water temperature, and condensate water temperature corresponding to each set of environmental conditions and circulating cooling water system operation scheme, and establish a system operation water temperature database; S2-2. Determine the three deep learning hyperparameters of Bayesian optimization: network depth, initial learning rate, and gradient descent momentum. The change of network depth is achieved by adjusting the number of convolution modules. The form of convolution modules is set as: convolution layer-normalization layer-activation layer. S2-3. Deep learning completes the training of all sample data, and finally uses the validation set error as an indicator to judge the training effect. The absolute value of the difference between the water temperature prediction result and the actual value does not exceed 0.1℃ as the prediction qualification standard. The Bayesian optimization process takes the training set, validation set data, and optimized hyperparameters as input, and the prediction error of the convolutional neural network training result and the corresponding network structure as output. The optimization process is implemented with the minimum error rate of the test set as the goal. The Bayesian optimization objective function is expressed as: In the formula: num represents the number of qualified prediction standards; Y pre , Y val Respectively represent the predicted value and the actual value of the validation set; length represents the length operation of the matrix; S2-4. Finally, the black box model is obtained.

3. The method for optimizing the period operation of a circulating cooling water system based on the maximum net profit of a power plant according to claim 1, characterized in that: The fouling deposition rate model and the peeling rate model in the net fouling rate model of the condenser tube in S3 are as follows: According to Hasson's classical theory, with concentration difference as the driving force, the dirt deposition rate is expressed as: Where: β is the mass transfer coefficient from the main solution to the scale layer boundary, m·s -1 ; C F is the salt concentration in the flowing medium, kg·m -3 ; C f is the salt concentration at the interface between the fluid solution and the dirt, kg·m -3 ;k R is the reaction rate coefficient, m 4 kg -1 ·s -1 ; C s is the saturation concentration of sparingly soluble salts in the fluid, kg·m -3 ; n is the scaling reaction order. For reactions only related to anions and cations, n = 2; Rearrange the above formula to Where: d ti is the inner diameter of the pipe, m; C F,Me , C F,An are respectively the concentrations of scaling anions and cations, kg·m -3 ; u is the cooling water flow rate in the tube, m·s -1 ; ν is the kinematic viscosity of water, m 2 ·s -1 ; The subscript ref indicates reference, t is the circulating water temperature in the condenser, K, which is the average temperature of the condenser inlet and outlet water; K R0 is the reaction rate constant, m 4 kg -1 ·s -1 ; E is the activation energy of the reaction, J·mol -1 ; R is the molar gas constant, which is 8.314 J·mol -1 ·K -1 , t d is the surface temperature of the scale layer, K; Assuming that the temperature gradient of the scale layer is 0, the scale peeling rate equation is expressed as: m r '=83.2u 2.54 r f d c (r 3 (vg) 1 / 3 d f Where: f is the scale density, kg·m - 3;d c is the scale layer crystal diameter, m.

4. The method for optimizing the period operation of a circulating cooling water system based on the maximum net profit of a power plant according to claim 1, characterized in that: The total power generation revenue and various costs in the two-layer optimization model built by S4 are as follows: The derivation process of the total power generation revenue formula of the power plant is as follows: The general expression of the power generation of the steam turbine unit is: Where: b1, b2 are power correction coefficients; R s is the steam load rate; N rtur is the rated power of the steam turbine, kW; α r is the proportion of exhaust resistance loss of steam turbine; By integrating the above formula over time, we can get the total power generation revenue within the cycle, and use the rectangular method to simplify the integral formula and solve it: Where: T is the duration of the adjustment cycle, h; n tur is the number of running steam turbines, e is the electricity price, yuan·kW -1 ·h -1 ,For some peak load plants, the electricity sales prices are different at different time periods; The operating cost of the pump unit of the circulating cooling water system in period T is: Where: N p is the total operating power of the circulating cooling water system pump unit, kW, calculated by the following formula: Where: subscripts "cs", "vf", and "ma" represent fixed speed pumps, variable frequency pumps, and water supply pumps, respectively; n is the number of pump units in operation; H is the pump head, m; η p , η tm , η m , η f are the efficiencies of the water pump, transmission device, motor, and inverter respectively; The formula for calculating the supplementary water cost is as follows: C ma =3600TW ma r ma Where: r ma The unit price of supplementary water, Yuan·m -3 ; The calculation formula for the pump adjustment loss cost is as follows: C ad =(|Δn vf |+|Δn cs |)p ad Where: Δn is the difference between the number of running pumps in this regulation cycle and the number of running pumps in the previous regulation cycle; p ad The start-up and shutdown cost of a single water pump unit, RMB; For power plants equipped with a rubber ball online cleaning system, the cleaning cost is expressed as follows: Where: N rp 、n rp They are respectively the operating power and the number of operations of the rubber ball pump, T rp is the running time of the rubber ball pump, h; num clean is the number of cleanings in this cycle, p r This is the additional cost for a single cleaning of the rubber ball, RMB.

Images